Meetingly
AI-powered video meeting platform where AI agents join live calls and automatically produce meeting summaries, searchable transcripts, and contextual post-call chat.
Live: https://meetingly-a.vercel.app
Repo: https://github.com/vanshikhaBhardwaj07/Meetingly
Features
- Real-time AI agents — custom agents join live video calls via the OpenAI Realtime API and hold speech-to-speech conversations during the meeting.
- Automated summaries — an event-driven Inngest pipeline ingests the call transcript after each meeting and generates a structured summary with GPT-4o.
- Searchable transcripts — full transcript search with highlighted matches.
- Post-call chat — a context-aware assistant grounded in the meeting summary, the agent's original instructions, and recent conversation history.
- Subscription tiers — free-tier usage limits (agents/meetings) enforced through tRPC middleware, with upgrades handled via Polar billing.
- Authentication — email/password plus Google and GitHub OAuth via Better Auth.
Tech Stack
Frontend: TypeScript, Next.js 15 (App Router), React 19, Tailwind CSS v4, shadcn/ui
API: tRPC, Zod, TanStack Query
Database: PostgreSQL (Neon, serverless), Drizzle ORM
Real-time: Stream Video SDK (WebRTC), Stream Chat, OpenAI Realtime API
Background jobs: Inngest
Auth: Better Auth (email/password, Google OAuth, GitHub OAuth)
Payments: Polar
Deployment: Vercel
Getting Started
Prerequisites
You'll need accounts/API keys for: Neon (Postgres), Stream (Video + Chat), OpenAI, Google & GitHub OAuth apps, and Polar.
Setup
npm install
Create a .env file with:
DATABASE_URL=
BETTER_AUTH_URL=http://localhost:3000
NEXT_PUBLIC_APP_URL=http://localhost:3000
GOOGLE_CLIENT_ID=
GOOGLE_CLIENT_SECRET=
GITHUB_CLIENT_ID=
GITHUB_CLIENT_SECRET=
NEXT_PUBLIC_STREAM_VIDEO_API_KEY=
STREAM_VIDEO_SECRET_KEY=
NEXT_PUBLIC_STREAM_CHAT_API_KEY=
STREAM_CHAT_SECRET_KEY=
OPENAI_API_KEY=
POLAR_ACCESS_TOKEN=
Push the database schema:
npm run db:push
Run the dev server:
npm run dev
Open http://localhost:3000.
Video call webhooks (transcripts, recordings, summaries, chat) require a public URL. Use npm run dev:webhook (ngrok) alongside npm run dev and point your Stream webhook endpoint at the tunnel's /api/webhook URL.
Deployment
Deployed on Vercel. Push to main to deploy; pull requests get their own preview deployments automatically.
